Twenty-one patients with schizophrenia who met criteria for neuroleptic treatment resistance or intolerance participated in a crossover, placebo-controlled, double-blind comparison of long-term typical neuroleptic and clozapine treatment. Clozapine significantly reduced total as well as positive and negative symptoms in comparison with both fluphenazine and placebo. Of the 21 patients, eight (38%) showed clozapine superiority on the basis of prospective response criteria. High levels of extrapyramidal side effects during fluphenazine treatment and later onset of illness were clinical predictors of clozapine superiority. Clozapine and fluphenazine equally reduced plasma homovanillic acid levels in comparison with placebo, although fluphenazine but not clozapine increased plasma prolactin level. A striking biologic difference between clozapine and fluphenazine was clozapine's enhancement of indexes of noradrenergic activity. Superior clozapine response was predicted by low ratios of cerebrospinal fluid homovanillic acid to 5-hydroxyindoleacetic acid, consistent with the notion that balance between dopaminergic and serotoninergic systems is important for clozapine's mechanism of action.